# Estimating LLM Consistency: A User Baseline vs Surrogate Metrics

arXiv: https://arxiv.org/abs/2505.23799 · HTML: https://mlsanity.com/paper/2505.23799/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 22 of its 59 references that are in the map.
- First posted on arXiv 26 May 2025.

## Builds on (most cited first, 22 of 22)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://mlsanity.com/paper/2203.11171/) (2022)
- [BERTScore: Evaluating Text Generation with BERT](https://mlsanity.com/paper/1904.09675/) (2019)
- [Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles](https://mlsanity.com/paper/1612.01474/) (2016)
- [A large annotated corpus for learning natural language inference](https://mlsanity.com/paper/1508.05326/) (2015)
- [Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks](https://mlsanity.com/paper/2003.01690/) (2020)
- [CoQA: A Conversational Question Answering Challenge](https://mlsanity.com/paper/1808.07042/) (2018)
- [Universal Sentence Encoder](https://mlsanity.com/paper/1803.11175/) (2018)
- [LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset](https://mlsanity.com/paper/2309.11998/) (2023)
- [Semantic Entropy Probes: Robust and Cheap Hallucination Detection in LLMs](https://mlsanity.com/paper/2406.15927/) (2024)
- [Membership Inference Attacks against Language Models via Neighbourhood Comparison](https://mlsanity.com/paper/2305.18462/) (2023)
- [A Survey of Confidence Estimation and Calibration in Large Language Models](https://mlsanity.com/paper/2311.08298/) (2023)
- [LLMs instead of Human Judges? A Large Scale Empirical Study across 20 NLP Evaluation Tasks](https://mlsanity.com/paper/2406.18403/) (2024)
- [LUQ: Long-text Uncertainty Quantification for LLMs](https://mlsanity.com/paper/2403.20279/) (2024)
- [Large Language Models are Not Yet Human-Level Evaluators for Abstractive Summarization](https://mlsanity.com/paper/2305.13091/) (2023)
- [Uncertainty quantification in fine-tuned LLMs using LoRA ensembles](https://mlsanity.com/paper/2402.12264/) (2024)
- [Optimizing Temperature for Language Models with Multi-Sample Inference](https://mlsanity.com/paper/2502.05234/) (2025)
- [LLM Whisperer: An Inconspicuous Attack to Bias LLM Responses](https://mlsanity.com/paper/2406.04755/) (2024)
- [Consistency in Language Models: Current Landscape, Challenges, and Future Directions](https://mlsanity.com/paper/2505.00268/) (2025)
- [Monte Carlo Temperature: a robust sampling strategy for LLM's uncertainty quantification methods](https://mlsanity.com/paper/2502.18389/) (2025)
- [Predicting the Performance of Black-box LLMs through Follow-up Queries](https://mlsanity.com/paper/2501.01558/) (2025)
- [HumanRankEval: Automatic Evaluation of LMs as Conversational Assistants](https://mlsanity.com/paper/2405.09186/) (2024)

## Built on by (most cited first, 8 of 8)

- [STED and Consistency Scoring: A Framework for Evaluating LLM Structured Output Reliability](https://mlsanity.com/paper/2512.23712/) (2025)
- [Controlling Distributional Bias in Multi-Round LLM Generation via KL-Optimized Fine-Tuning](https://mlsanity.com/paper/2604.05756/) (2026)
- [User Perceptions vs. Proxy LLM Judges: Privacy and Helpfulness in LLM Responses to Privacy-Sensitive Scenarios](https://mlsanity.com/paper/2510.20721/) (2025)
- [CryptoQA: A Large-scale Question-answering Dataset for AI-assisted Cryptography](https://mlsanity.com/paper/2512.02625/) (2025)
- [Security and Privacy Prompts in the Wild: What Users Ask LLMs and How LLMs Respond](https://mlsanity.com/paper/2606.18062/) (2026)
- [Every Wrong Answer Counts: Option-Level Psychometrics for LLM Multiple-Choice Benchmarks](https://mlsanity.com/paper/2608.02966/) (2026)
- [Harmonizing Diverse Models: A Layer-wise Merging Strategy for Consistent Generation](https://mlsanity.com/paper/2510.14915/) (2025)
- [Evaluating Multimodal LLMs across Text and Audio Modalities for Accessible Disaster Assistance](https://mlsanity.com/paper/2608.14651/) (2026)

## Abstract (arXiv)

Large language models (LLMs) are prone to hallucinations and sensitive to prompt perturbations, often resulting in inconsistent or unreliable generated text. Different methods have been proposed to mitigate such hallucinations and fragility, one of which is to measure the consistency of LLM responses -- the model's confidence in the response or likelihood of generating a similar response when resampled. In previous work, measuring LLM response consistency often relied on calculating the probability of a response appearing within a pool of resampled responses, analyzing internal states, or evaluating logits of responses. However, it was not clear how well these approaches approximated users' perceptions of consistency of LLM responses. To find out, we performed a user study (n=2,976) demonstrating that current methods for measuring LLM response consistency typically do not align well with humans' perceptions of LLM consistency. We propose a logit-based ensemble method for estimating LLM consistency and show that our method matches the performance of the best-performing existing metric in estimating human ratings of LLM consistency. Our results suggest that methods for estimating LLM consistency without human evaluation are sufficiently imperfect to warrant broader use of evaluation with human input; this would avoid misjudging the adequacy of models because of the imperfections of automated consistency metrics.